Will develop the machine learning methods for PBSHM, building in domain knowledge to move diagnostic capability between structures. Cutting-edge machine learning methodologies based on Transfer Learning will be developed to ensure optimal knowledge transfer between structures.
Will research the enabling technology for real-world application of the Theme 1 algorithms, moving beyond traditional database technology. It will develop and extend PBSHM using (hyper-)graph and complex network theory. New approaches to security will instil trust in industrial partners. A novel risk/utility-based approach to PBSHM will be developed, allowing cost-benefit analyses and decision support.
Will create the knowledge delivery systems which will populate the Framework with rich and diverse information from real structures and physics-based models. It will develop novel low-cost instrumentation, highly tuned for the field: low-power and edge computing-supported. It will develop new technology for physics-based modelling and physics-informed machine learning that will combine data-based and model-based SHM in an effective fusion tailored to populations.
Is a comprehensive programme of requirements analysis and expert elicitation in partnership with industrial partners, leading to very large-scale demonstrators linked into the ICs. In particular, bridges are a very heterogeneous group of multi-scale multi-material structures and will be a challenge to the new PBSHM theory and Framework. This WP will integrate PBSHM into existing Infrastructure Management Systems (IMSs).